Skip to main content
Version: 1.2

SpEL Cheat Sheet

Expressions and types

Expressions used in Nussknacker are primarily written using SpEL (Spring Expression language) - simple, yet powerful expression language. SpEL is based on Java (reference documentation), but no prior Java knowledge is needed to use it.

The easiest way to learn SpEL is looking at examples which are further down this page. Some attention should be paid to data types, described in more detail in the next section, as depending on the context in which data are processed or displayed, different data type schemes are in use.

Check out SpEL overview for the overview of how SpEL is used by Nussknacker.

Data types and structures

The data types are used primarily for:

  • validation - e.g. to detect attempt to use incorrect data type, for example numeric field instead of a string, or checking if field used in expression exists at all.
  • code completion - suggestions appearing in UI when editing expressions.

Types of events in the Kafka streams or data returned by enrichers can be often discovered from some sort of schema registry, for example Confluent Schema Registry, SQL table schema or description of REST API. Nussknacker can also infer types of variables defined by user.

The data types used in the execution engine, SpEL expressions and data structures are Java based. These are also the data type names that appear in code completion hints. In most cases Nussknacker can automatically convert between Java data types and JSON and AVRO formats. JSON will be used for REST API enrichers, while AVRO should be first choice for format of Kafka messages.

Below is the list of the most common data types, with their JSON and Avro counterparts. In Java types column package names are omitted for brevity, they are usually java.lang (primitives), java.util (List, Map) and java.time

Basic (primitive data types)

Java typeJSONAvroComment
nullnullnull
StringstringstringUTF-8
Booleanbooleanboolean
Integernumberint32bit
Longnumberlong64bit
Floatnumberfloatsingle precision
Doublenumberdoubledouble precision
BigDecimalnumberbytes or fixed + decimalenable computation without rounding errors
UUIDstringstring + uuiduuid

More information about how to declare each type in Avro you can find in Avro ducumentation, especially about Avro logical types.

Records/objects

In Nussknacker, the following data types share common processing characteristics:

  • object in JSON
  • record or map in Avro
  • Map and POJO in Java

In many cases Nussknacker can convert between them automatically. For the user, the most significant difference is (using Avro terminology) between record and map. Both can describe following JSON structure:

input = { name: 'John', surname: 'Doe'}

The main difference is that in case of record Nussknacker "knows" which fields (name and surname) are available and suggests and validates fields and their types. For example, #input.name is valid, while #input.noname or #input.name > 0 as field name or type do not match.

On the other hand, map describes "generic" structure - Nussknacker tacitly assumes it can contain any field of any type.

Nussknacker usually infers structure of record from external source (e.g. AVRO schema), but it can also detect it from map literals.

Arrays/lists

In Nussknacker (e.g. in code completion) JSON / Avro arrays are refered to as Lists; also in some context Collection can be met (it's Java API for handling lists, sets etc.).

Handling date/time.

Date/time data types

Formats of date/time are pretty complex - especially in Java. There are basically three ways of storing date:

  • as timestamp - absolute value, number of milliseconds since 1970-01-01T00:00:00 UTC. In Nussknacker this is usually seen as Instant or Long. This format is handy for storing/sending values, a bit problematic when it comes to computations like adding a month or extracting date.
  • as date/time without timezone information (this is usually handy if your system is in one timezone). Converting to timestamp is done using Nussknacker server timezone. In Nussknacker they are usually represented as LocalDate and LocalDateTime. Suitable for date computations like adding a month or extracting date.
  • as date/time with stored timezone. In Nussknacker usually seen as ZonedDateTime. Suitable for date computations like adding a month or extracting date.
  • as date/time with stored time offset. In Nussknacker usually seen as OffsetDateTime. Contrary to ZonedDateTime doesn't handle daylight saving time. Quite often used to hold timestamp with additional information showing what was the local date/time from "user perspective"

The following table mapping of types, possible JSON representation (no standard here though) and mapping of AVRO types (int + date means int type with date logical type):

Java typeJSONAvroSampleComment
LocalDatestringint + date2021-05-17Timezone is not stored
LocalTimestringint + time-millis or long + time-micros07:34:00.12345Timezone is not stored
LocalDateTimestringnot supported yet2021-05-17T07:34:00Timezone is not stored
ZonedDateTimestringlong + timestamp-millis or timestamp-micros (supported only in sinks)2021-05-17T07:34:00+01:00
OffsetDateTimestringlong + timestamp-millis or timestamp-micros (supported only in sinks)2021-05-17T07:34:00+01:00[Europe/Paris]
Instantnumberlong + timestamp-millis or timestamp-micros2021-05-17T05:34:00ZTimestamp (millis since 1970-01-01) in human readable format
Longnumberlong, long + local-timestamp-millis or local-timestamp-micros123456789Raw timestamp (millis since 1970-01-01)

Conversions between date/time types

Conversions of different types of dates are handled either by

  • #DATE helper methods e.g.:
    • #DATE.toEpochMilli(#zondeDate)
  • methods on some types of objects, e.g.
    • #instantObj.toEpochMilli returns timestamp for #instantObj of type Instant
    • #localDate.atStartOfDay() - returns LocalDateTime at midnight for #localDate of type LocalDate
    • #localDateTime.toLocalDate - truncates to date for #localDateTime of type LocalDateTime
    • #zonedDate.toInstant - converts ZonedDateTime to Instant
    • #instant.atZone('Europe/Paris') - converts ZonedDateTime to Instant
    • #instant.atOffset('+01:00') - converts OffsetDateTime to Instant
  • automatically by implicit conversion mechanism e.g.
    • #instant.atZone('Europe/Paris') - Europe/Paris String was automatically converted to ZonedId
    • #instant.atOffset('+01:00') - +01:00 String was automatically converted to ZonedId
    • #time.isAfter('09:00') - 09:00 String was automatically converted to LocalTime
    • #date.isBefore('2020-07-01') - 2020-07-01 String was automatically converted to LocalDate
    • #dateTime.isAfter('2020-05-01T11:00:00') - 2020-05-01T11:00:00 String was automatically converted to LocalDateTime

Date/time utility methods

DATE helper contains also some other useful helper methods, mainly for date range checks and computations of periods and durations e.g.:

  • #DATE.isBetween(#localTime, '09:00', '17:00') - checks if LocalTime is in (inclusive) range <09:00, 17:00>
  • #DATE.isBetween(#dayOfWeek, #DATE.MONDAY, #DATE.FRIDAY) - checks if DayOfWeek is in (inclusive) range <MONDAY, FRIDAY>
  • #DATE.isBetween(#localDate, '2020-06-01', '2020-07-01') - checks if LocalDate is in (inclusive) range <2020-06-01, 2020-07-01>
  • #DATE.isBetween(#localDateTime, '2020-06-01T11:00:00', '2020-07-01T11:00:00') - checks if LocalDateTime is in (inclusive) range <2020-06-01T11:00:00, 2020-07-01T11:00:00>
  • #DATE.periodBetween(#from, #to).getMonths - computes Period between from and to and return number of full months between those two dates
  • #DATE.durationBetween(#from, #to).toDays - computes Duration between from and to and return number of full days between those two dates. Keep in mind that Duration is not daylight saving time aware - it computes seconds difference and divide it by number of seconds in given period.
  • In case of days it will be 86400 seconds.

Some useful constants are also available:

  • #DATE.MONDAY, #DATE.TUESDAY, ... - day of weeks
  • #DATE.JANUARY, #DATE.FEBRUARY, ... - months
  • #DATE.zuluTimeZone - Zulu timezone which always has time zone offset equals to UTC
  • #DATE.UTCOffset - UTC offset
  • #DATE.defaultTimeZone - Default time zone for Nussknacker application

Parsing of date/time

Also, #DATE_FORMAT helper methods can be used to parse or format certain data type from/to the String. It is not recommended to use parsing in scenarios because it will obfuscate logic. Better way is to configure properly message schema. But sometimes it is the only way to handle it. Available helpers:

  • #DATE_FORMAT.parseOffsetDateTime('2020-01-01T11:12:13+01:00') - parse OffsetDateTime in ISO-8601 format
  • #DATE_FORMAT.parseOffsetDateTime('2020-01-01T11:12:13+01:00', 'yyyy-MM-dd'T'HH:mm:ssXXX') - parse OffsetDateTime in given DateTimeFormatter format
  • #DATE_FORMAT.parseOffsetDateTime('2020-01-01T11:12:13+01:00', #dateTimeFormatter) - parse OffsetDateTime using given DateTimeFormatter

Equivalent variants of parse methods are available also for other date/time types: LocalTime, LocalDate, LocalDateTime, Instant and ZonedDateTime.

Formatting of date/time

To format date/time can be used #DATE_FORMAT.format(#dateTime) method which accept various date/time types and formats it in ISO-8601 format. Also DateTimeFormatter can be used directly via e.g. #DATE_FORMAT.formatter('EEEE').format(#date). Other formatter factory methods:

  • #DATE_FORMAT.formatter('EEEE', 'PL') - creates DateTimeFormatter using given pattern (EEEE) and locale (PL)
  • #DATE_FORMAT.lenientFormatter('yyyy-MM-dd') - creates lenient version of DateTimeFormatter using given pattern. Lenient parser may use heuristics to interpret inputs that do not precisely match format e.g. format E will accept: mon, Mon and MONDAY inputs. On the other hand, formatter created using #DATE_FORMAT.formatter() method will accept only Mon input.
  • #DATE_FORMAT.lenientFormatter('yyyy-MM-dd EEEE', 'PL') - creates lenient version DateTimeFormatter using given pattern and locale

For full list of available format options take a look at DateTimeFormatter api docs.

SpEL syntax

Literals

Most of the literals are similar to JSON ones, in fact in many cases JSON structure is valid SpEL. There are a few notable exceptions:

  • Lists are written using curly braces: {"firstElement", "secondElement"}, as [] is used to access elements in array
  • Empty record is {:}, to distinguish it from empty list: {}
  • Strings can be quoted with either ' or "
  • Field names in records do not to be quoted (e.g. {name: "John"} is valid SpEL, but not valid JSON)
ExpressionResultType
'Hello World'"Hello World"String
truetrueBoolean
nullnullNull
{1,2,3,4}a list of integers from 1 to 4List[Integer]
{john:300, alex:400}a map (name-value collection)Map[String, Integer]
#inputvariable

Arithmetic Operators

ExpressionResultType
42 + 244Integer
'AA' + 'BB'"AABB"String

Conditional Operators

ExpressionResultType
2 == 2trueboolean
2 > 1trueboolean
true AND falsefalseboolean
true && falsefalseboolean
true OR falsetrueboolean
true || falsetrueboolean
2 > 1 ? 'a' : 'b'"a"String
2 < 1 ? 'a' : 'b'"b"String
#nonNullVar == null ? 'Unkown' : 'Success'"Success"String
#nullVar == null ? 'Unknown' : 'Success'"Unknown"String
#nullVar?:'Unknown'"Unknown"String
'john'?:'Unknown'"john"String

Method invocations

As Nussknacker uses Java types, some objects are more than data containers - there are additional methods that can be invoked on them. Method parameters are passed in parentheses, usually parameter details are shown in code completion hints.

ExpressionResultType
'someValue.substring(4)"Value"String
'someValue'.length()9Integer

Accessing elements of a list or a record

ExpressionResultType
{1,2,3,4}[0]1Integer
{jan:300, alex:400}[alex]a value of field 'alex', which is 400Integer

Filtering lists

Special variable #this is used to operate on single element of list. Filtering all the elements uses a syntax of .?. In addition to filtering all the elements, you can retrieve only the first or the last value. To obtain the first element matching the predicate, the syntax is .^. To obtain the last matching element, the syntax is .$.

ExpressionResultType
{1,2,3,4}.?[#this ge 3]{3, 4}List[Integer]
#usersList.?[#this.firstName == 'john']{'john doe'}List[String]
{1,2,3,4}.^[#this ge 3]{1}Integer
{1,2,3,4}.$[#this ge 3]{4}Integer

Mapping lists

Special variable #this is used to operate on single element of list.

Examples below assume following structure:

listOfPersons: List[Person]
person1 = name: "Alex"; age: 42
person2 = name: "John"; age: 24
listOfPersons = {person1, person2}
ExpressionResultType
{1,2,3,4}.![#this * 2]{2, 4, 6, 8}List[Integer]
#listOfPersons.![#this.name]{'Alex', 'John'}List[String]
#listOfPersons.![#this.age]{42, 24}List[Integer]
#listOfPersons.![7]{7, 7}List[Integer]

Safe navigation

When you access nested structure, you have to take care of null fields, otherwise you'll end up with error. SpEL provides helpful safe navigation operator, it's basically shorthand for conditional operator: #someVar?.b means #someVar != null ? #someVar.b : null

ExpressionResultType
null.someFieldjava.lang.NullPointerExceptionjava.lang.NullPointerException
null?.someFieldnullNull

Invoking static methods

It is possible to invoke Java static methods directly with SpEL. Nussknacker can prevent invocations of some of them due to security reasons. Invoking static methods is advanced functionality, which can lead to incomprehensible expressions, also code completions will not work with many of them. If you need to invoke the same method in many places, probably the best solution is to create additional helper.

ExpressionResultType
T(java.lang.Math).PI3.14159..Double

Chaining with dot

ExpressionResultType
{1, 2, 3, 4}.?[#this > 1].![#this > 2 ? #this * 2 : #this]{2, 6, 8}Double

Type conversions

SpEL has many built-in implicit conversions that are available also in Nussknacker. Mostly conversions between various numeric types and between String and some useful logical value types. Some examples:

Input valueInput typeConversion target type
12.34DoubleBigDecimal
12.34fFloatBigDecimal
42IntegerBigDecimal
42LLongBigDecimal
'Europe/Warsaw'StringZoneId
'+01:00'StringZoneOffset
'09:00'StringLocalTime
'2020-07-01'StringLocalDate
'2020-07-01T'09:00'StringLocalDateTime
'en_GB'StringLocale
'ISO-8859-1'StringCharset
'USD'StringCurrency
'bf3bb3e0-b359-4e18-95dd-1d89c7dc5135'StringUUID

You can also use explicit conversions that are available in utility classes and build-in java conversion mechanisms:

ExpressionResultType
#NUMERIC.toNumber('42')42Number
#NUMERIC.toNumber('42').toString()'42'String
#DATE_FORMAT.parseOffsetDateTime('2018-10-23T12:12:13+00:00')1540296720000OffsetDateTime
#DATE_FORMAT.parseLocalDateTime('2018-10-23T12:12:13')2018-10-23T12:12:13+00:00LocalDateTime
'' + 42'42'String

Built-in helpers

HelperFunctions
GEOSimple distance measurements
NUMERICNumber parsing
CONVGeneral conversion functions
DATEDate operations (conversions, useful helpers)
DATE_FORMATDate formatting/parsing operations
UTILVarious utilities (e.g. identifier generation)
AGGAggregator functions